SEReDeEP replaces ReDeEP's lexical scores with semantic entropy probe scores, but its claimed 3-10% accuracy gains are contradicted by its own tables and its probes are trained on the evaluation datasets.
Deepseek-r1 release, 2025
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
SEReDeEP: Hallucination Detection in Retrieval-Augmented Models via Semantic Entropy and Context-Parameter Fusion
SEReDeEP replaces ReDeEP's lexical scores with semantic entropy probe scores, but its claimed 3-10% accuracy gains are contradicted by its own tables and its probes are trained on the evaluation datasets.